THEMIS 5.0 at the DIGISET Forum: Building Trustworthy AI for Democratic Participation

As part of the New European Bauhaus Festival, THEMIS 5.0 was proud to be represented at the DIGISET Forum, a satellite event hosted by CollectiveUP and FARI – AI for the Common Good Institute in Brussels. Bringing together experts from across the fields of artificial intelligence, social economy, civic participation and innovation, the forum explored how AI can contribute to more inclusive, sustainable and democratic societies.

Representing THEMIS 5.0, Adam Doulgerakis of Athens Technology Center (ATC) joined the panel discussion Democracy, Participation and the New European Bauhaus, alongside Clare Twelvetrees from The Guardian Foundation and Imre Keserű from Mobilise VUB, with moderation by Barna Petrányi of Pro Progressione. The discussion focused on a timely question: how can democratic participation evolve in the age of AI, and what role should technology play in shaping more inclusive decision-making processes?

Adam Doulgerakis

AI as a Support Layer, Not a Substitute for Democracy

Drawing on his work leading the THEMIS 5.0 media pilot and co-creation activities in the news media sector, Adam highlighted a central principle of the project: AI should augment human decision-making rather than replace it.

While AI has the potential to lower barriers to participation by making information more accessible, supporting multilingual engagement, and helping people navigate complex issues, these benefits are not automatic. AI systems can also reinforce existing inequalities, obscure important assumptions, or privilege voices that are already more visible in digital spaces if they are not designed carefully.

As Adam noted during the discussion, trustworthy AI is not only a technical challenge but also a social one. Questions of who defines success, who is affected by decisions, who can challenge outcomes, and whose voices are absent from the process are just as important as algorithmic performance.

This perspective aligns closely with the vision of the New European Bauhaus, which seeks to combine sustainability, inclusion and beauty while respecting local contexts and community values. AI can support these ambitions only when it strengthens meaningful participation rather than merely creating the appearance of it.

Trust Requires More Than Technical Performance

A key theme of the panel was trust. Within THEMIS 5.0, trustworthiness is understood through a range of characteristics including accuracy, robustness, fairness, explainability and accountability. While reliable technical performance is essential, Adam emphasised that technical excellence alone is insufficient to build public trust.

Explainability emerged as a particularly important factor. People should not only see the outputs generated by an AI system; they should also understand how those outputs were produced, what evidence informed them, what limitations exist, and how the system influenced the final decision.

Crucially, explainability must be inclusive. An explanation that only technical experts can understand does little to build trust among citizens, journalists, policymakers or other stakeholders. Different users require different forms of explanation depending on their role, expertise and needs. Human-centred explainability is therefore a cornerstone of the THEMIS 5.0 approach to trustworthy AI.

From Participation to Influence

The discussion also addressed the relationship between AI and democratic participation.

AI can help broaden participation by making it easier for more people to engage with complex information and contribute their perspectives. However, meaningful democracy requires more than collecting inputs. It requires influence.

Participants must be able to see how their contributions are interpreted, considered and reflected in decision-making processes. Without transparency and traceability, AI-enabled participation risks becoming a superficial exercise that gathers feedback without demonstrating its impact.

For this reason, the importance of contestability and accountability was emphasised. People must be able to question AI-generated outputs, challenge interpretations, provide additional context and understand how their contributions affect outcomes. Trust grows when individuals can see not only that they were heard, but also how their voices shaped decisions.

A Shared Responsibility

The panel also explored broader concerns about the values embedded within AI systems. AI technologies often appear neutral, but they inevitably reflect assumptions present in data, design choices, institutional priorities and governance structures.

One of the key lessons emerging from THEMIS 5.0 is that concepts such as fairness, robustness and accuracy do not have universal definitions. Their meaning depends on context, stakeholders and societal priorities. As a result, communities should be involved not only as users of AI systems but also as active participants in defining the values, risks and trade-offs that shape them.

Without this involvement, AI risks imposing a narrow set of assumptions while appearing objective.

Looking Forward: AI for Democratic Participation

The discussion at the DIGISET Forum reinforced a message that:

AI can make participation broader, faster and more accessible. But democracy also requires deliberation, disagreement, context and accountability. The goal is not to automate participation, but to design AI systems that help people deliberate better and demonstrate that their voices have real consequences.

We would like to thank Liliana Carillo, CollectiveUP, FARI – AI for the Common Good Institute, DIGISET, and all the organisers and participants for creating a valuable space to discuss the future of AI, democracy and citizen participation.

As THEMIS 5.0 continues its work across the media, healthcare and port sectors, these conversations remain essential to ensuring that AI systems are not only innovative, but also transparent, inclusive, accountable and worthy of public trust.